Overview:
We are planning a 2 node cluster. The Active node will be 8 GB RAM with one
instance of SQL 2005 and 150 GB of SAN disk. This is considered the
production node with failover to the second node (Passive).

The passive node also has 8 GB of RAM, 150 GB of SAN disk, and has 4 local
test instances of SQL 2005, and the failover production instance. We think
that we will need to limit the 4 instances to 5 GB of RAM (Local Instance 1 -
1.5 GB, Local Instance 2 - 1.5 GB, Local Instance 3 - 1 GB, Local Instance 4
- 1GB). This will leave us 2 GB for the failover instance and 1 GB for the
OS. For extended outages, we would like to be able to shutdown the local
instances of SQL and allow the failover production instance to access more
than the 2GB reserved.

Questions:
- On the passive node, are there issues having local instances of sql and
allowing for a failover instance? What SQL 2005 memory settings are
available to allow this configuration?

- On the passive mode, in the above configuration, can the local instances
each point to their own folders in the shared logical drive?

- Are their white papers describing best practices?

Thanks,

don.ramler (AT) discussions (DOT) microsoft.com

Comments Inline


"Don.Ramler" <don.ramler (AT) discussions (DOT) microsoft.com> wrote

nodes. The Cluster service arbitrates ownership and control. LUNS are
mapped to clustered disk resources and are specific to each SQL instance.
production and development/test systems.
instance of SQL. You can create a LUN and present it only to the second
node. Do not create it as a clustered disk resource and it will look like a
locally attached disk. The you can use subfolders for each instance.

--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP

Geoff,

Are there any memory setting guidelines with regard to the multiple instance
passive node?

Are there any white papers on configurations of this sort?
--
Don Ramler


"Geoff N. Hiten" wrote:

I don't know of any right away. As for "active" and "passive" nodes, all
nodes are equivalent so the choice of current host is arbitrary. You can
set it as a preference. Using a normally inactive node as a host for other
instances is not recommended, so I doubt there are any references on how to
do that.

--
Geoff N. Hiten
Senior Database Administrator
Microsoft SQL Server MVP




"Don.Ramler" <don.ramler (AT) discussions (DOT) microsoft.com> wrote

The services are supposed to start automatically on node 2. Are all of the
dependent resources online - disks, IP, network name? SQL won't come up
unless all dependencies are online. It may be because a disk got hit in the
power failure.

BTW, this is a programming newsgroup. You'll get better luck if you post to
the clustering group. (I've cross-posted this there.)

--
Tom

----------------------------------------------------
Thomas A. Moreau, BSc, PhD, MCSE, MCDBA, MCITP, MCTS
SQL Server MVP
Toronto, ON Canada
https://mvp.support.microsoft.com/profile/Tom.Moreau


"NC3" <ncoleman3 (AT) yahoo (DOT) com> wrote

After experiencing a power outage sql server services on node one of
our active/passive cluster failed to start and failed over to node 2
as expected. I tried starting the services on node 2 without success.
Is this by design or should the services be able to start. We
configured node 2 to have it's services manually started which we were
able to do fortunately for me, should I confiure the services on node
one to start manually or leave them to automatic start.

If the services should be started and they are feeling can anyone tell
me what to look for to get the services started again.

Thanks in advance

NC3

Your DBA is not correct. Failover is automatic.

--
Tom

----------------------------------------------------
Thomas A. Moreau, BSc, PhD, MCSE, MCDBA, MCITP, MCTS
SQL Server MVP
Toronto, ON Canada
https://mvp.support.microsoft.com/profile/Tom.Moreau


"GJL" <GJL (AT) discussions (DOT) microsoft.com> wrote

Hi NC3,

I'm not going to be much help, but maybe it will help a little. I don't
know the command off hand, but our dba has explained to me before that you
have to run a command on the passive node to promote it to active before it
is ready.

Sorry I'm not much more help...

~g

"NC3" wrote: